My chatbots need to be trained when we get new data or feedbacks from users. Can someone provides ways how these chatbots can learn on themselves and become intelligent day by day?
Some of techniques may be :
- Reinforcement learning - though I could not find a successful implementation and have asked a separate question. Any successful implementation reference would be helpful.
- Building a pipeline with a user feedback - this has 2 drawbacks. One user or admin has to provide the correct answer. Second, a single feedback data may not be enough to retrain the bot.
Any expert advice on - self optimizing chatbot architecture OR some code reference OR some paper trying to solve this problem (a paper with code may be)?